Block DCT-based robust watermarking using side information extracted by mean filtering

A novel watermarking method is presented based on the concept of communications with side information. We investigate the characteristic of mean filtering to formalize the problems of watermark embedding and extraction. This leads to a non-additive embedding strategy and a sophisticated blind detection mechanism. With regard to geometric attacks (without cropping), block-based watermarking and moment normalization are both addressed without relying on pilot signals. The new watermarking scheme has been verified by Stirmark and a new benchmark.